Most LED display problems are fixed in minutes with a screwdriver and the spares box — because LEDs are the component least likely to fail. The killers, in order, are power supplies, water, connectors, and configuration. Below: a three-minute triage, a twelve-row diagnostic table to bookmark, and each failure explained with the bench evidence that names it.

It is Saturday, 6:40 AM, doors at 9, and a stripe of your screen is dark. This guide exists for exactly this moment. I have answered the Saturday-morning calls for a decade, run the teardown bench at Eyecatchmedia, and kept the failure statistics the whole time. When owners call about LED display problems, they blame the LEDs first. The log says otherwise.

Most screen problems are power problems wearing a costume.

The rule for today: panic fixes nothing; sequence does. Work the triage, find your symptom in the table, then read that problem’s entry — each one includes what to tell the factory on the call, so the support conversation takes five minutes instead of forty.

The 3-Minute Triage (Before Anything Else)

1. Power cycle in the right order: screen → processor → source. Wait 30 seconds. Restart in reverse. Half of all “dead” calls end here.

1. Check the obvious with your eyes: breaker on, cables fully seated, status LEDs on the sender and receiving cards. A cleaner sharing a socket trips more screens than lightning does.

1. Photograph the symptom before touching anything. Timestamped photos are what turn a warranty conversation from your word into a claim.

The Diagnostic Table (Bookmark This)

The Service Log Behind the Table

One statistic, and it is ours, drawn from ten years of service records: the LED itself is the most reliable part on the wall. Power supplies fail first. Water is second. Connectors third. Configuration fourth — someone re-addressed a card, or a sender lost its settings overnight, and the screen takes the blame. Keep that order in mind. It sorts the twelve problems faster than any datasheet does.

1. A single dead pixel or module

At the bench, a dead pixel is one of two things: the lamp failed, or the solder joint under it opened. Both name themselves under a loupe — a dark lamp face, or a hairline crack at the pad after a cold winter’s thermal cycling.

The fix is a swap, not a repair.

Nobody reflows one SMD lamp for a living. By the time you have the module out, you are most of the way to fitting a whole tile, and a replacement module costs less than the repair labor would. That is why modules exist. Single-pixel repair is not economical, and pretending otherwise just buys you a second failure at the same address. One five-minute check first: press gently near the dead pixel. If it flickers back, the ribbon seat was the culprit, not the lamp — reseat it and close the cabinet.

2. A section of the screen goes dark

A clean half of the wall out, or a vertical band dead, points at the data chain, not the power chain. The half-screen signature tells you which link died. The first receiving card drives roughly half the wall, so a clean half means that card or its feed from the sender. A band a cabinet or two wide means the hub card behind that band.

Swap diagnostic, ten minutes: exchange the two receiving cards, or borrow the hub feed from a healthy neighbor. If the dark section moves with the card, the card is the patient. If the dark section stays put, the card is innocent and the ribbon or hub port owns the fault. Swap the cheap parts before you diagnose the expensive ones — that sentence is most of any honest led screen repair guide.

3. The whole screen is black: led display not working at all

Work the power chain backward, in this order.

Breaker first. It is embarrassing how many “dead” walls were tripped by a cleaner sharing a socket. Then the power supplies: fan spinning, indicator LED lit? Then measure the output rails at the drive connector — anything near 4.8 volts or better means the supply lives; a sagging rail means the supply is the patient. Last, the receiving card status LED, which separates “no power” from “power but no data.”

And trust your nose. What does a dying power supply smell like? Vented electrolytic capacitors have a signature — hot varnish sweetness, sweet and burnt at once, unmistakable after you have smelled it once. The taxi-top screen on our bench last spring announced its diagnosis before we opened the case: both supplies had vented, and the cooked boards confirmed it. (That teardown, and what taxi-top screens actually cost to run, sits in our taxi-top cost teardown.)

4. Vertical dead stripes after rain

Water ingress has a signature, and it is not subtle. Open the rear door after a wet night and shine a light along the bottom edge of the modules. White-green corrosion mapping to the bottom of the cabinet — worst at the lowest point, where water pooled — is the entry path drawn for you in oxidation. That is the two-minute rear-view diagnosis, and it has ended more arguments with more suppliers than any spec sheet.

The water usually moves through the ribbon connectors, wicking and condensing until the pins grow a white crust, and the stripes above the crust die in vertical columns because that is how the ribbon feeds them. On the bench we see two kinds of cabinets: glue-sealed ones with a cracked edge-bead that let water in slowly and never out, and gasketed profiles with drainage weeps at the bottom that expect water and give it a door. If your cabinet has no weep holes, that is your answer about the cabinet.

5. Rolling bands that only show up on camera

Dark bands crawl through the footage, but the wall looks fine in person. That is refresh rate below roughly 2,000 Hz fighting the camera shutter.

Eyes forgive. Cameras don’t.

There is no firmware fix. Scan rate is bought at the driver IC stage; the economy IC will band in every filmed close-up forever, and no settings menu changes that. Picking the right scan spec is a purchasing decision, not a settings one, and our P3.91 rental LED screen for the UK runs high-refresh drivers as standard for exactly this reason. Moiré is the different cousin — a geometry problem between pixel pitch, sensor, and shooting distance. Moiré moves when the camera moves; refresh banding does not. Different disease, different cure.

6. Flicker you can see with the naked eye

Visible flicker is electrical before it is logical. Two causes account for nearly everything on our bench: grounding loops and PSU ripple.

Put a scope on the supply’s output. A healthy supply idles near 50 mV of ripple. The tired one in our worst flicker complaint read 300 mV, and every module behind it was bathing in that noise. Six times out of ten the fix is the ground, not the parts: isolate the screen’s ground from lighting dimmers and rigging, and the flicker leaves with it. Four times out of ten the trace stays fat, and the supply goes in the scrap bin.

7. Colors drifting between cabinets

One cabinet sits a touch magenta against its neighbors, and it was not like that last year. Usual culprit: bin drift. LEDs get sorted into wavelength bins at the factory, and a replacement module — or a phase-two expansion — from a different batch lands on the wall with a different color floor.

This is why we hold matched-batch reserves for every wall we ship and print the batch number on the cabinet label. Recalibration buys you time; a good calibrator can pull a mixed-batch wall into visible tolerance. But re-binning is the real fix, and it only exists if somebody kept the batch records. Ask for them at delivery, not two years later when the wall has a patchwork look and the original supplier has stopped answering email.

8. The whole screen dims over months

Lamp decay. Every LED fades; the question is how fast. Our teardown measurements, lamp by lamp under the integrating sphere: quality lamps lose under 8% brightness a year, no-name lamps lose 20 to 30%. At 30% a year, a wall rated 5,000 nits is a 3,500-nit wall inside two years, and the owner swears it “was brighter at delivery.” It was.

Check the claimed nits against our brightness guide before you buy, not after. Then do one thing at handover: get a baseline luminance report. Without that baseline, a decay claim is your word against a datasheet. With it, the claim takes ten minutes and a light meter to prove, and the warranty clause finally has teeth.

9. Ghosting and afterimage

Trails behind moving content come from two places: cable quality and termination on long ribbon runs, or driver ICs on their way out.

Swap-test the ribbon first. It is the cheap part and the abused part — pinched by cabinet seams, stretched by re-hangs, cooked by the PSU mounted next door. A known-good ribbon clears the ghost, done in five minutes. If the ghost survives the ribbon swap, the module’s driver IC is the suspect and the module goes back under warranty. Long runs are where termination matters most: a forty-cabinet wall built with mismatched ribbons will ghost in ways a ten-cabinet wall never will.

10. No signal on the screen

The chain is sender, cable, receiver — check it in that order, and know that in our call log, 90% of no-signal calls die at the cable.

Sender card status lights first: no input detected is a source problem, input detected but no output is downstream. Then the cable: reseat both ends, then replace it. CAT5/6 between sender and first receiving card has real length limits — past roughly a hundred meters you want fiber, not hope. Then there is HDCP, which kills more live events than any hardware fault: a source and a processor that refuse to handshake will black the screen while every status light glows green. Drop the source’s output resolution once to test it. Half the “processor failures” I have seen were handshake standoffs.

11. The screen restarts at random

A supply running at 95% load in a sealed cabinet gets hot, its protection trips, and it takes itself offline to survive. Minutes later it comes back like nothing happened, and the owner has a restart log that looks like a poltergeist. Undervoltage under load behaves the same way — the wall browns out mid-show and reboots in front of the client.

The rule from the factory side: derate to 70% of rated load. A 300-watt supply drives 210 watts, maximum. Yes, that costs more up front, and purchasing hates it every budget round. It also means the supply never lives at its thermal limit, and it is the difference between a five-year PSU and a ten-year one. If your restarts cluster on hot afternoons, you already have the diagnosis; confirm it by aiming a fan at the cabinet vents and watching the restarts stop.

12. Dead on arrival after shipping

Transport kills connectors. Road vibration works ribbons and power connectors half a millimeter out of their seats, and half a millimeter is enough to kill a column of data or an entire cabinet.

First move after delivery: open the back and reseat every ribbon and power connector. Twenty minutes for a typical wall, and it has rescued more “DOA” screens than any other single act on this page. Check the packing too — cabinets that shifted in transit took their interconnects with them.

This is why burn-in happens before packing and reseating happens after transit. Our burn runs hold 48 to 72 hours and pull 2–3% of early failures — cheap deaths at the factory instead of expensive ones at your venue.

The spares box: what to hold, and how much

The spares box is the cheapest insurance in this business. Ours ships with every wall, and here is the list with quantities:

• Spare modules at 5% of the wall, weighted to the faces that fail — bottom rows and the weather side, per the corrosion map in problem 4

• One power supply per 20 cabinets

• Two receiving cards and two hub cards, matching what is already on the wall

• Ribbon cables, exact length and connector, three of each

• The magnet tool, if the wall is front-service

A front-service wall without the magnet tool is a cabinet you cannot open from the correct side. We learned that the honest way. A spares box weighted evenly across all faces is half a box; the bottom of the wall does the dying.

Prevention, from the factory floor

Four specs do most of the preventing, and each one kills specific problems from the list above:

• PSU derating to 70% — kills #3, #6, and #11, and roughly doubles supply life

• Gasketed waterproofing with drainage weeps — kills #4, the most expensive failure a screen can suffer

• 48–72 hour burn-in pulling 2–3% — kills #1 and #2 before they ever reach your venue

• Front service — does not prevent #1 or #2, but shrinks them from a rig day to a five-minute module swap

Ask any supplier which of the four they actually do. The answers sort the market faster than the datasheets do.

LED display problems: quick answers

Can dead pixels be repaired? Not as such. A dead pixel is a failed lamp or an opened solder joint, and single-pixel repair is not economical at any scale we have ever run. The real LED screen dead-pixel fix is a module swap — five minutes on a front-service cabinet, one spare module out of the 5% box.

Why is my LED screen flickering on camera? Refresh rate. Below about 2,000 Hz the panel fights the camera shutter and the footage bands. The wall can look flawless to the eye while the recording is unusable. Scan rate is set at the driver IC stage, so it is a purchasing decision, not a settings decision.

How do I know if a power supply is failing? Random restarts that cluster on hot days, visible flicker, output rails that sag under load, and the hot-varnish smell of vented capacitors. Any two of those together and I replace the supply rather than argue with it.

Why is my screen dimmer than the day it was installed? Lamp decay, most likely accelerated by cheap bins. Compare against the delivery luminance report; if the drop outruns 8% a year on a named-brand lamp, that is a warranty conversation, and the report is what makes it winnable.

What should an on-site spares kit contain? Modules at 5% weighted to the failing faces, one PSU per 20 cabinets, receiving and hub cards, ribbons, and the magnet tool. Weighted to the faces that fail, not spread evenly — the bottom of the wall does the dying.

The Warranty Call Script

When the fix above doesn’t close it, the call goes better with the right material ready:

• Photographs of the symptom, wide and close, timestamped — plus one of the open cabinet rear

• The batch number from the cabinet label; it pulls your wall’s burn-in log and lamp batch records in minutes

• What you already tried, in the table’s vocabulary: “rails measure 4.6V under load,” “dark section follows the receiving card,” “ripple reads 300 mV after grounding was isolated.” Specific words shorten the call; “it’s broken” lengthens it

• What the 3-year warranty covers: parts — modules, power supplies, receiving cards — with replacements per contract terms. What it doesn’t: lightning, vandalism, unauthorized modification, and water ingress through a cabinet opened in the rain

Keep this page as your LED screen repair guide; the twelve fixes above close out most of the calls we take. And when it is time to specify rather than repair — screens we build ship with 5% spares, front service, burn-in logs, and engineers who answer the Saturday call. See our P3.91 rental LED screen for the UK or tell us what the wall needs to survive: contact us